By Philip Keitel

Recent economic events have reminded us that accurately predicting a particular consumer’s credit risk can be a difficult task. Anticipating risk among consumers with thin or no credit histories is even trickier. Yet with estimates of the size of the thin- and no-file population ranging from approximately 35 million to 70 million consumers (depending on how these groups are defined), there is much discussion about the potential use for alternative data in credit scoring this segment.

In this context, “alternative payment data” generally refers to payment behavior not ordinarily gathered in the current consumer credit reporting and scoring processes. These include recurring payments, such as utility payments, payments to telecommunications providers, or insurance payments; use of non-traditional financial services, such as check cashing, payday loans, certain money remittance services; and use of particular payment instruments, such as prepaid cards.

Even Fair Issac Corp., the company that created the industry-standard FICO scoring algorithms, has developed a new scorecard using alternative data. Early tests show that its Expansion Score is able to obtain a score for the thin- or no-file consumers who have little or no credit data on record up to 90 percent of the time.

That being said, use of alternative data in credit scoring is still evolving and has yet to be widely adopted. However, recent research indicates that alternative payment data can be effective in predicting risk among these thin-and no-file consumers, fostering the notion that, some day, alternative sources of information will lead to an improved ability to build credit scores. If further research confirms these preliminary findings, credit scoring systems that incorporate alternative data could be better predictors than traditional methods of default and delinquency risks among this population.

However, a number of challenges remain. These include distinguishing between types of alternative payments data that may predict risk better than others, resolving challenges remaining on both the demand and supply sides of the current marketplace, proving to a greater degree the usefulness of this data in credit-risk modeling, and negotiating an assortment of operational issues. To discuss research and challenges surrounding the use of alternative payments data in credit scoring, the Payment Cards Center of the Federal Reserve Bank of Philadelphia recently hosted a workshop with Arjan Schütte, associate director of the Center for Financial Services Innovation.

The findings presented during that workshop are summarized as follows.

Risk and Red Flags

One of the first challenges in developing the market for alternative data is determining which data are best predictors of risk. Issuers and financial institutions can benefit from analyzing alternative data along three specific dimensions.

First, alternative payments are examined on a transactional level: Is the underlying transaction “cash-like” or “credit-like?” The fundamental premise is that the more credit-like a transaction is, the more helpful it will be in determining the likelihood that a thin-or no-file consumer will make future payments on traditional credit products.
The second dimension relates to how widely a particular type of alternative payment is made by the population at large. The more widely conducted or adopted an instrument, the better. The underlying assumption is that if use or adoption is extensive, data analysis standards can be efficiently applied across a large population. Otherwise, if use is limited, the incremental benefit associated with gathering the data may be less than needed to justify the costs associated in doing so.

Along the third dimension, alternative payments data should be considered in relation to data furnishers. The underlying premise is that if data furnishers are highly concentrated, scale efficiencies increase, making it more likely that those furnishers’ efforts to report this information will be successful.

Based on this framework, utility and telecommunications payments as two types of credit-like alternative payment transactions have characteristics that would allow for a broad application of analytical frameworks at a potentially reasonable cost.

A Matter of Supply and Demand

The existing supply of alternative payments data is another matter. Data furnishers, one of three types of organizations on the supply side of the alternative data market (the other two being data repositories and data scoring firms), supply alternative data to repositories that warehouse and manage the databases storing the data. However, data furnishing is rapidly evolving, with many new and well-established companies contributing data or analytics to scoring methods that incorporate at least some elements of alternative data in order to improve underwriting decisions. However, further develop­ment of the market depends largely upon there being a ready supply of data. Many believe that supply will increase only when it is clearly demonstrated that there are strong economic benefits derived from using alternative data, such as improved screening of new applicants and lower default rates.

While the supply of alternative data may indeed expand as the cost-benefit trade-offs become clearer, those on the demand side of the market may need to offer some form of monetary incentive to data furnishers to encourage them to share full-file payment data. Nonetheless, as more research is undertaken to determine the impact of using alternative payments data on risk analytics, and as lenders show more interest in better evaluating risk for underserved customers, the supply side will continue to mature.

On the demand side, lender interest hinges on a number of factors. First, data sources must be broad and deep, offering redundancy and high hit rates. Second, data delivery systems should be improved so that they integrate both alternative and traditional data and allow lenders to use existing channels and sources. Third, risk managers must build trust with new systems by realizing incremental benefits gained by incorporating alternative data into credit underwriting and other business decision models. Essentially, lender demand will grow as it becomes evident that incorporating alternative data into credit scoring models will also allow the profitable expansion of portfolios.

In Practice: The Use of Alternative Data

To better understand the potential that alternative data may hold, researchers have undertaken several studies to analyze whether alterna­tive data can be shown to improve estimates of default and delinquency risk. One such study, conducted by the Policy and Economic Research Council and the Brookings Institution Urban Markets Initiative, examined more than 8 million credit records, containing at least one alternative trade line reported on a full-file basis — that is, both positive and negative data. These files included thin- and no-file consumers. The study found that the addition of even one alternative trade line, relative to either a utility payment or a telecommunications payment, proved useful and showed improved predictive power for several commercial scoring models.

However, this result was primarily tied to an ability to newly score those individuals who couldn’t previously be scored because they had either thin-or no-credit files, rather than an improvement in predicting risk for those who already had credit scores based solely on traditional data.

Separate studies carried out by risk-modeling firms showed that by using alternative data, these firms were able to score a significant portion of loan applicants without a traditional credit score, and effectively measure risk among thin-and no-file applicants. Under Fair Isaac’s FICO Expansion Score, LexisNexis’ Risk Review, L2C’s Link2Credit or First Score Direct products, 70 to 100 percent of the sample applicants without traditional credit scores could be effectively scored by incorporating alternative data. Based on the subsequent actual payment performance, these tests allowed the firms to analyze the effectiveness with which alternative data credit scores predicted future default and delinquency rates at the time of the loan application. This affirmed that alternative data strengthen a lender’s ability to predict risk associated with thin-and no-file borrowers and thereby facilitate lending to lower-risk borrowers in this consumer segment that otherwise might be denied credit.

From Sidelines to Mainstream

Without a doubt, incorporating alternative payments data in current credit scoring practices will be challenging. If deciding which alternative data to incorporate into data scoring models is not difficult enough in itself, a number of additional factors will affect the development of the market for alternative data, including the costs of modifying legacy systems; the costs and complexities of changing IT infrastructures; legal and regulatory hurdles; broad economic impacts of extending the market for consumer credit; and various other challenges. Partnering with a third-party provider can help mitigate these challenges.

Indeed, the continued evolution of supply and demand for alternative data in the credit information markets will itself center on crucial factors, such as whether there is a business case motivating furnishers of alternative data to voluntarily share payment information with data repositories and to what extent the data can help to better predict default and delinquency risk for those consumers with thin-and no-credit files. Nevertheless, for information furnishers, data repositories, data scoring firms, lenders and consumers alike, there seem to be real social and economic incentives for incorporating alternative data into lending decisions by traditional financial institutions.

Ultimately, if using alternative payments data in credit scoring improves the accuracy of existing models, and the market for alternative data develops, there is the potential for a significant, long-term overall impact on consumer credit in America. For thin-and no-file consumers, many of whom have traditionally fallen outside the financial mainstream, improvements in credit scoring represent the prospect of joining the mainstream and the opportunity to obtain traditional financial products and services. For lenders, improvements in credit scoring represent the opportunity to soundly and securely expand their business.

About the Author

Philip Keitel is an industry specialist in the Payment Cards Center of the Federal Reserve Bank of Philadelphia. This article has been adapted from a Federal Reserve Bank of Philadelphia Payment Cards Center Discussion Paper authored by Julia Cheney. For more information on the Payment Cards Center or to access a complete version of the original paper, please see www.philadelphiafed.org/pcc. To learn more about the Center for Financial Services Innovation, please visit www.cfsinnovation.com/ about.php.